| Filename | Latest commit message | Latest commit date |
|---|---|---|
23dc0939e0
There is no Linux MsQuic binary in any NuGet package, and the v2.5.9 GitHub release carries only test bundles, so the Windows script's approach does not transfer. What does exist is Microsoft's Debian 13 pool, which is the same distribution Yggdrasil runs, and the msquic source tree at the matching tag for the POSIX headers the NuGet omits. `scripts/build-quic-native.sh` fetches the pinned `libmsquic_2.5.9_amd64.deb` and the three headers `msquic.h`, `msquic_posix.h` and `quic_sal_stub.h`, and verifies all four by SHA-256 the same way `build-quic-native.ps1:14` verifies the NuGet. A digest that stops matching is a changed dependency, not a warning. Two details the deb forced. It carries three library files and no licence text, so the MIT licence is copied from the one this repo already committed for the Windows plugin — the same text for both platforms. And it ships only `libmsquic.so.2.5.9`, with no `libmsquic.so` for `-lmsquic` to resolve against, so the script makes that link in the extracted tree rather than in the image. The Linux CMake branch links `-fvisibility=hidden`, so only the twelve v2 exports leave the `.so` and the Windows-only v1 five are absent by construction rather than by discipline, and sets `RPATH $ORIGIN` so the `libmsquic.so.2` shipped beside the bridge resolves without being installed system-wide. `build-quic-native.ps1`'s default output directory becomes `artifacts/quic-native/win32-x64` so the two platforms' outputs sit beside each other instead of on top of each other. Both existing callers — `build-unity-package.ps1:80-82` and the Native.Tests csproj `:20-22` — pass an explicit `-OutputDirectory`, so neither is affected. Built and smoked in the workstation's Docker `debian:13` engine pinned by digest: g++ 14.2.0, glibc 2.41, `libgamecult_mesh_quic_native.so` 85,456 bytes with `NEEDED libmsquic.so.2` and `RUNPATH $ORIGIN`, and the same 15-check v2 FFI smoke that runs on Windows passes under Node 24 on that image. That is a Linux build on a Linux host and nothing more: Q6 ruled the release artifact comes from the Actions job, and this one is the development loop. A Linux host still needs MsQuic's own runtime dependencies present — libssl3t64, libnuma1, libxdp1, libnl-route-3-200 — which is now written down, because the first smoke run failed on a missing libxdp.so.1 and the reason was not obvious from the error. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> |
||
| .. | ||
| CMakeLists.txt | ||
| cultmesh_quic_native.cpp | ||
| README.md | ||
GameCult.Mesh.Quic.Native
One MsQuic body for every CultMesh host that is not .NET: two roles (provider and consumer), two platforms (Windows x64 and Linux x64), and no callbacks across the host boundary.
The bridge owns MsQuic registration, configuration, listener, connection and
stream handles, TLS credential loading, the cultmesh-state-v1 ALPN, the
1-byte-kind plus 4-byte-LE-length QUIC stream framing, and one event queue per
runtime. It owns transport and nothing else: it never decodes a CultMesh frame,
never decides delivery semantics, never coalesces, never reconnects, and in the
v2 path never decides whether to trust a certificate. Frame meaning stays in
GameCult.Mesh and in the TypeScript codec; trust stays with the host.
No host callbacks
Every crossing is the host calling in, including the one blocking wait. MsQuic's worker threads only ever push onto the queue and wake that wait, so no host runtime is called from a thread it does not own. That is what lets one binary serve a Node process through FFI and a Unity player through P/Invoke without either one's threading rules leaking into the other.
It is also why nothing here includes a Node header, a Unity header, or a JavaScript engine: the bridge does not know what is on the other side.
Two ABIs
v2 is the runtime-neutral queue API, and the one new hosts should use:
cultmesh_quic_runtime_open / runtime_close
cultmesh_quic_listener_open / listener_close
cultmesh_quic_connection_open / connection_certificate_complete / connection_shutdown
cultmesh_quic_stream_open / stream_send_frame / stream_shutdown
cultmesh_quic_next_event
cultmesh_quic_last_error
cultmesh_quic_event is a fixed 64-byte struct described by layout rather than
by a shared header, so a host declares it in its own FFI. Its type is one of:
1 listener_new_connection, 2 connection_connected, 3
connection_certificate_received (payload: the DER certificate), 4
connection_shutdown (payload: a UTF-8 reason), 5 stream_started
(stream_kind taken from the stream's first byte), 6 stream_frame (payload:
one whole encoded frame, length prefix stripped), 7 stream_send_complete, 8
stream_shutdown, 9 listener_stopped.
cultmesh_quic_next_event returns 0 on timeout, 1 with the event filled and its
payload copied, 2 when the payload buffer is too small — the event stays at the
head of the queue and out_required is set, so the host can allocate and ask
again — and negative on a bad call.
A client connection defers certificate validation: the bridge indicates the
certificate, emits event 3 carrying the DER, and answers MsQuic only when the
host calls cultmesh_quic_connection_certificate_complete. If the host never
answers, MsQuic's 10-second handshake timeout closes the connection and event 4
reports it. Providers load their certificate as PKCS12; the advertised
cert-sha256 pin is computed by the host from that same PKCS12, not here.
v1 is the five exports the shipped Unity package imports by name
(cultmesh_quic_open, _state, _poll, _error, _close). They are Windows
only, matching CultMeshNativeQuicRealtimeTransport.CanConnect, and are
implemented on top of a private v2 runtime holding one connection. They decide
nothing the v2 path does not, with one exception: a v1 connection carries a
SHA-256 pin that is compared here, in C, because that is the contract the
managed connector was built against. A v2 connection never carries one.
Building
Windows x64, from the repo root:
powershell -File scripts/build-quic-native.ps1
MsQuic comes from a pinned NuGet package verified by SHA-256, and the build uses
MSVC with /Brepro so a rebuild reproduces the committed Unity plugin DLL.
Linux x64, on a Debian 13 host:
scripts/build-quic-native.sh
MsQuic comes from Microsoft's Debian 13 pool as a pinned .deb, and the POSIX
headers from the msquic source tree at the matching tag; all four downloads are
verified by SHA-256. The bridge is linked with -fvisibility=hidden so only the
v2 exports leave the .so, and with RPATH $ORIGIN so the libmsquic.so.2
shipped beside it resolves without being installed system-wide.
A Linux host still needs MsQuic's own runtime dependencies present:
libssl3t64, libnuma1, libxdp1, libnl-route-3-200.
Where an artifact was built is part of what it is. The Linux binary is linked against Debian 13's glibc, so a build from anywhere else is a different artifact whatever the file name says.